Torsional vibrations in power transmission shafts are produced by load fluctuations or by fluctuations of the power source or transmission system. Detection of these vibrations, especially if they can be isolated from other vibrations and noise, is a means of early detection of flaws in the transmission system. The smaller the fluctuation which can be detected, and the more specific the location of the source, the earlier such flaws can be identified so that potentially serious problems can be averted.
Indirect means of monitoring, such as microphones or accelerometers--mounted for example to a bearing block--can provide a warning when failure is imminent, but the information of interest in their signals is partially masked by ordinary vibrations and cannot identify the specific source of vibrations. Previous optical means of detecting torsional vibrations are either not sensitive enough to detect small fluctuations, or are expensive, fragile laboratory devices not suitable for continuous monitoring in the hostile environment of a power system. Additionally, many of these systems must be installed before assembly of the power system and/or require access to the ends of shafts in the system.
It has been found that detailed analysis of torsional variations, for example those that recur at the same point during a rotational cycle of a shaft or system of shafts can be used to identify wear or damage to gears and other transmission components used to transmit power to and receive power from the shaft. For example, a gear with a worn tooth will result in uneven application of power to the shaft due to delay in engagement of the worn tooth with its mating tooth. Prior art systems do not permit the desired regular monitoring and detection of very small recurrent variations in power transfer, and may lack the sensitivity to identify developing problems before they have already become audible or visible to an experienced mechanic. It would therefore be desirable to provide a system and method capable of resolving and reporting very small torsional vibrations which could be permanently installed on an oceangoing vessel or other mode of transportation, or in an industrial facility.
Hartman and Palmer, in ISA Transactions, Volume 12, pp. 186-190 (1973), disclose application of an optical technique to measurement of torsional oscillations on a non-rotating beam in a laboratory and were able to measure a resolution of 10.sup.-4 arc-seconds. In the field, however, the inventors have found that resolution is limited by motion of optics components relative to each other, and with a moving shaft resolution will also be affected by shaft run-out and other shaft vibrational modes. Resolution of much less than one arc-second under these dynamic conditions appears difficult to reach by any optical technique due to vibration of the optics.
An article by J. M. Vance and R. S. French titled "Measurement of Torsional Vibration in Rotating Machinery" presented at the Design Engineering and Technical Conference of ASME, Oct. 7-10, 1984, shows the use of optically transduced signals from a special tape applied to a shaft.
U.S. Pat. No. 4,551,017 to Mannava et at. shows a system which is said to optically measure torsional vibration of a shaft using a photodetector which senses the passage of a grating on the moving surface of the shaft. An optional reference beam at the same location is used to detect other motion of the shaft. The photodetector output is passed through a zero-crossing detector. The system disclosed uses a laser Doppler velocimetry system for illuminating the grating. Mannava et at. use a non-zero Bragg diffraction order from the shaft mounted grating to obtain a Doppler frequency shift due to the surface velocity of the shaft. The collected first order diffracted light is then heterodyned with a reference beam. This approach requires coherent incident and receiving optics and requires that the light collection angle be carefully chosen to avoid zero and second diffraction orders, and further requires that the vector normal to the shaft surface remain constant, and that the position of the plane tangent to the shaft surface remain constant. Further, it requires that a diffraction grating be machined into the surface of the shaft, or that a grating, if suitable grating exists, be attached to the surface of the shaft.
U.S. Pat. No. 4,995,257 to Leon shows an angular shaft vibration monitor with two optical sensors spaced along the length of the shaft. Beams of light are shined toward the shaft, and are reflected by markings on the shaft, with the scattering of light detected by optical sensors. The difference between the sensor signals is used to measure torsional vibration. However, the markings used are single marks. While such systems may be effective in detecting large random torsional vibrations, they are incapable of finely resolving higher frequency torsional variations occurring only at certain points over an entire revolution of a shaft.
U.S. Pat. No. 4,433,585 to Levine shows a method and apparatus for measuring torsional deviation of a shaft using optical diffraction transmission gratings. Signals from two photodiodes are passed through comparators and their phases compared. The resulting signal is low-pass filtered to generate a signal representing torsional deviation. In this system, the diffraction gratings are arranged at opposite ends of the shaft, making it difficult or impossible to retrofit an existing industrial or naval shaft for such detection. Such a system cannot be applied on multiple shafts or on shafts with inaccessible ends. On very large shafts, the size of the optical disc required becomes an obstacle to easy installation, and the large optical discs will introduce error into measurements because of their own vibrations. Finally, Levine does not disclose a system which detects variations in torsional deviation, i.e. torsional vibration. It is desirable to monitor not simply the steady state twist of the shaft, but oscillations in the shaft twist, principally the first few torsional pendulum modes. In the largest shafts, the twist due to the steady state load is already small, perhaps 50 arc-seconds/foot. Oscillation amplitudes will generally be a small fraction of this amount, say 5 arc-seconds/foot. Even this represents a very high energy oscillation for which monitoring would be desirable.
U.S. Pat. No. 3,885,420 to Wolfinger shows a method and apparatus for measuring torsional vibration of a shaft. A signal from a shaft rotation sensor is passed through a zero crossing detector and its phase is compared with a second signal. The resulting signal is low-pass filtered to generate a signal representing torsional vibration. However, the sensors used are proximity sensors used with gear teeth. These can only resolve low frequency, large amplitude variations. Furthermore, the signals are processed by a phase-locked-loop and low pass filtered, limiting the application to less than one octave of shaft speed range and to low frequency variations. The system disclosed is applied to a constant speed 60 hz power generator for which it is suited, but like the Mannava et al. system described above, systems of this type do not provide true torsional vibration readings and are incapable of resolving either small amplitude or high frequency vibrations.
U.S. Pat. No. 4,317,371 to Wolfinger shows the same non-optical torsional vibration monitor to which quadrature has been added along which separate frequency bandpass filters to help discriminate against periodic electrical interference which masks the signal. Another patent to Wolfinger et al., U.S. Pat. No. 3,934,459, shows another system using multiple nonoptical sensors in a complex machine to detect variations in torsional oscillation between plural rotating shafts.
U.S. Pat. No. 5,001,937 to Bechtel et al. shows an optically based torsion sensor using an adhesive band applied to the shaft. The band has alternating high and low reflectivity regions, the passage of which is sensed to detect torsion. Like the U.K. patent to Lowman, this disclosed system uses a light source to produce a single spot on the target shaft at two or more locations. The ability to resolve shaft twist appears to depend upon reducing the spot to a small size, in particular much smaller than the line spacing. An example in the disclosure of Bechtel et al. resolves only three minutes of arc. Lines are produced by painting or machining the shaft.
U.K. Patent Application GB 2 093 991 A to R. J. Lowman shows a torque measurement apparatus with axially spaced alternating reflective and non-reflective areas, the passage of which is detected with a pulsed laser light beam directed to a single spot on each of the two axially spaced reflective and non-reflective areas on the shaft. The reflective areas are produced by grinding or polishing, and the non-reflective areas by painting or etching.
U.S. Pat. No. 4,641,027 to Renner et al. shows an apparatus for measuring static or dynamic torque which optically measures relative movement of a flexible surface to make the torque measurement. This system uses lines which at zero torque are parallel to the axis of the shaft to measure shaft torque. The lines are strained as the shaft is strained, and thus their angular orientation is slightly changed. The system disclosed uses the angle of the lines as indicated by the spacing of Moire fringes between two transparent cylinders or disks to measure the torsional strain of a shaft.
U.S. Pat. No. 4,806,454 to Barraud et al. and U.S. Pat. No. 4,997,747 to Yoshida et al. show generally the production of diffraction gratings using photo processes. However, these patents are directed primarily to etching techniques and do not show or suggest application of grating lines to a cylindrical shaft. A very large number of lines is required to cover the perimeter of any shaft greater than about 2.5" in diameter. Because of the number of lines and the need for precise spacing, diffraction errors and lens aberrations inherent in conventional techniques (such as photoreduction of a larger pattern onto a shaft) make it impossible to apply these techniques to forming very fine lines about the circumference of a shaft.
Photoelectric torque transducers, such as those used in the 8800 series Motor evaluators and 1038 Motorized Dynamometer made by Vibrac Corporation, are also known, but are not suitable for torsion vibration measurement because frequency response is limited by line spacing and by the digital counter processing method used. Such systems do not have the precision to resolve under 300 arc-seconds. These transducers were designed to include a shaft and torsion rod which must transmit the shaft loads, and have maximum loads of 200 lb-ft.
Similarly, prior laser doppler velocimetry systems developed by Brtiel & Kjaer, Dantee, and TSI, Inc. are laboratory research instruments unsuitable for fleet deployment due to their expensive construction, fragile nature, and difficulty of use. Also, as far as the inventors are aware, these systems have not been used with very finely spaced grid lines applied to the shaft, which the inventors have found improves accuracy and ease of use.
In conclusion, none of the prior art systems known to the inventors provides an entirely satisfactory apparatus and method which can be permanently retrofitted to existing industrial and shipboard applications to continuously detect and analyze torsional vibrations occurring in rotating shafts.